A health system's AI book

Context

A health system weighs twelve initiatives spanning clinical (scribe, sepsis warning, imaging) and administrative (prior-auth, revenue-cycle coding, denials). Clinical value is real but slower to prove; admin value is faster and fundable now.

The decision

Revenue-cycle coding and denials management scale now — proven and high risk-adjusted ROI; the imaging-triage and bed-management pilots are cut — high spend, unproven, and the clinical-safety bar makes the ramp long.

What most miss

Clinical prestige biases the book toward glamorous pilots. The risk-adjusted math usually says the unglamorous admin automation funds the clinical ambition.

Stakes

Mis-allocating toward slow clinical pilots starves the admin automations that actually pay for the program.